Bulgarian Evidential Forms and Wierzbicka’s NSM

inferential, and mirative forms, incorporating Wierzbicka’s universal semantic primitives know and think. Our aim is to provide a stepping stone towards precise intra-and inter-lin-guistic comparisons.


Introduction
The impetus for this paper came from a publication on the semantics of evidentiality by the eminent Polish-Australian linguist Anna Wierzbicka.In Slavic studies, this category is discussed mostly with reference to Bulgarian and Macedonian, where the category has grammaticalised.Existing uniquely as a grammatical category amongst the other Slavic languages, the so-called re-narration (sensu largo) has been the focus of interest in the fields of typology, contact linguistics, areal linguistics, and, significantly, semantics.
In Bulgarian linguistics, the study of evidentiality becomes particularly intense from the last two decades of the 20th century onwards.This period sees the publication of an extensive seminal study by Georgi Gerdžikov (Герджиков, 1984), important contributions by Ivan Kucarov and Ruselina Nicolova (Куцаров, 1984; Ницолова, 1984), as well as work published outside Bulgaria most notably by Zlatka Guentchéva and Jean-Pierre Desclés (e.g.Guentchéva & Desclés, 1982).The interest in evidentiality continues in later decades and is studied both by Bulgarian linguists, as well as outside of the boundaries of Bulgaria.
Work specifically in Bulgarian is complemented by contrastive studies.For instance, in the multi-volume Bulgarian-Polish contrastive grammar, and particularly in Koseska-Toszewa et al. (1996) and Korytkowska and Roszko (1997), evidentiality is discussed in the context of the so-called imperceptive modality, seeking to compare the two languages in terms of semantic, rather than formal categories with the help of tertium comparationis (Косеска-Тошева & Балтова, 2004, p. 11).Eftimova (Ефтимова, 2023) explores evidentiality from the perspective of contemporary media studies in relation to the (non-)credibility of information more generally.
In her publication on evidentiality, Wierzbicka (1996) takes a broader view and seeks to find the place of Bulgarian (and Macedonian)1 evidentiality in a wider typology of languages.In this book, Wierzbicka seeks to build a universal model of the invariant meanings of evidentiality by explicating them via the framework of Natural Semantic Metalanguage (NSM).NSM, developed by Wierzbicka over a number of years (for a relatively recent presentation see Goddard & Wierzbicka, 2014;Wierzbicka, 2014), is based on the claim that there is a set of basic universal semantic primitives.The primitives, which consist of words or other simple linguistic expressions present in any human language, are meant to supply a tool for a linguistic and cultural analysis able to pin down complex or culturally specific expressions or grammatical categories, as well as articulate culturally specific values via maximally transparent and universally translatable notions.Wierzbicka (1996) is a typological study, and so Bulgarian/Macedonian are placed alongside a number of typologically unrelated languages, amongst them Kashaya (a language of the Pomo people spoken in California), Quechuan (Peru, Bolivia), Wintu (Northern California), and Maricopa (Arizona).
For Bulgarian/Macedonian data and interpretations, Wierzbicka draws on Friedman (1986).Wierzbicka's analysis is therefore influenced by a particular view of the verbal system, and more specifically by a particular conception of the interaction between perfect and evidentiality.In this work, Friedman essentially backgrounds evidentiality as a separate 'formal' category, or at least his view of its nature and size makes it different from the category usually recognised in Bulgarian mainstream linguistics.We find several points where we disagree with the analysis, but a detailed discussion would take us too far afield, besides, we won't be able to say much that hasn't been said already in recent publications on evidentiality (Asenova & Markou, 2022;Нинова, 2014;Търпоманова, 2015).Although in his view a category of evidentiality does not exist, Friedman doesn't deny the existence of evidential meanings, but he sees them as a 'by-product', i.e. a semantic extension of the perfect.In his view, the perfect (called indefinite past) has to be understood in its opposition to tenses denoting anteriority like the aorist, or imperfect ('definite past').According to Friedman, the use of the definite past entails a commitment on behalf of the speaker to the truth of a given statement (it is confirmatory).In this respect, the meaning of the indefinite past (perfect) is unmarked.This is what makes it appropriate when talking about events of which the speaker has no first-hand knowledge, hence evidentiality as a potential meaning of the perfect.
Since this is the starting point of Wierzbicka's analysis, this part of it could be understood (indirectly) as an NSM analysis of past vs.perfect tense.In Wierzbicka's theory, all complex meanings in all languages can be represented as different configurations of a small number of basic concepts.think and know specifically play an important role in grammar.She proposes the primitives know and think as the obvious, intuitive, and universal semantic indefinibilia (undefinables), referring to Leibniz's 'alphabet of human thought' and Descartes's enduring thesis that the concept 'cogito' cannot be defined, or simplified further.know, for instance, can be used to distinguish statements from interrogatives: statements are based on the semantic component know, whereas interrogatives can be expressed via 'don't know -want to know'.They are also important for the meaning of (non-) evidentials, for which Wierzbicka formulates in advance the paraphrases 'know because I see', 'know because I hear', or 'think but don't know that I know', and so forth.know surfaces when the speaker subjectively commits to the truth of the statements they make.think, on the other hand, is used when there is uncertainty.know is also the basic element in the paraphrases describing non-indicative modal categories in the contrastive Bulgarian-Polish grammar.Although Friedman's (1986) understanding of evidentiality is not co-extensive with the description that has become established in Bulgarian linguistics, cf. for instance, the exposition in Nicolova (Ницолова, 2008), he does accept that some forms can be characterised as evidentials.Wierzbicka in turn conveys the distinction between indicative and evidential via the formulae 'I know this not because someone said something' and 'I think this because someone said something'.

Perfect: know or thi̇nk?
Following Friedman, Wierzbicka considers evidentiality (primarily in Macedonian) to be a semantic extension, a particular meaning, of a polysemic perfect.According to Friedman, but also Aronson, Penčev, Stankov, and others (see Герджиков, 2003, pp. 63-64 and references therein), the forms of the perfect (incl.the past perfect) are to be distinguished from the aorist, and imperfect based on the feature 'confirmatory'.This presumed lack of confirmation is the reason Wierzbicka settles on the semantic primitive think.Given that mostly the focus of her paper is on presumed perfect forms with evidential semantics, her proposal could be taken as a suggestion to use think as the semantic primitive for the Bulgarian (and Macedonian) evidential.2We defer the discussion of evidentiality to the next section.Here we want to briefly address the question of whether think can be used to model the non-evidential uses of the perfect.
We do so with the proviso that it is not our intention to address the more general question of the meaning of the perfect.Not only is this outside of the scope of our discussion, but it would also be impossible for us within this small space to do justice to the complexity of this topic, or to the extensive research others have already done.
We will confine ourselves to pointing out that the use of the perfect does not entail uncertainty, which is what is implicit in the primitive think.As is clear from (1a), the perfect can be used to report on the speaker's own past activities and perceptions, as well as communicate activities and perceptions of interlocutors or third persons (1b and 1c).
(1) a. Az sâm xodila na more mnogo pâti, no nikoga ne sâm viždala tolkova mnogo meduzi.'I have visited the seaside many times, but never have I seen so many jellyfish.'b.Ti si xodila v Krakow i znaeš, če ima mnogo hubav velikdenski bazar.'You have been to Krakow and know that there is a very nice Easter market there'.
c. Toj beše pokanil na partito vsičkite si prijateli.'He had invited all his friends to the party'.
Following the perfect with an explicit statement of uncertainty as in (2) sounds odd.
'He had invited all his friends to the party, but I find it difficult to believe'.
We therefore side with Gerdžikov, who contends that when using the perfect/past perfect forms, the speaker does confirm the veracity of a statement, though the confirmation may relate not to the action itself, but to its result.We therefore conclude that an NSM semantic formula for the perfect would include know, rather than think.
Our aim for the remainder of the paper would be to use semantic primitives from Wierzbicka's NSM metalanguage and established ways of combining them to discuss the meaning of the evidential.We will expand the discussion to include the dubitative.We will also give some attention to the inferential and, more tentatively, the mirative.Next, we will turn to the evidential.

The evidential: knowledge and certainty
In this section, we focus specifically on the semantics of the evidential, the form which occurs most frequently and has the greatest obligatoriness of use.
In the previous section, we concluded that the use of the perfect, and the indicative tenses more generally, corresponds in Wierzbicka's NSM to the primitive know.We accept know by default.Indeed, we would consider know to be the default.Wierzbicka's primitives code semantic meaning, but the fundamental nature of know is suggested also in pragmatics by Grice's Co-Operative Principle, more specifically the quality maxim: speakers undertake to say what they believe to be true and have no evidence to be false.This isn't to say that speakers cannot talk of things they believe not to be true, or of whose truth they cannot be sufficiently certain.Observance of the Co-operative Principle, however, would place them under obligation to provide linguistic markers when the default commitments to truth don't apply.The evidential is potentially one such marker of a non-default interpretation.Wierzbicka's formulae for Bulgarian suggest that the evidential expresses uncertainty (think), or uncertainty related to source of information ('think because someone said something').
That evidentiality expresses non-firsthand knowledge is well-established in the literature.Traditionally, scholars identify two types of sources of information that prompt its use: (i) specific, known, mentioned source of information, or (ii) unknown, collective knowledge.In (3), we illustrate the former.
'Yesterday morning the child went out to school very early, and yet didn't catch the bus.' b.Včera sutrinta deteto izleze za učiliste mnogo rano i pak ne xvanalo rejsa.'Yesterday morning the child went out to school very early, and yet (allegedly) didn't catch the bus.' c.Včera sutrinta deteto izljazlo za učiliste mnogo rano i pak ne xvanalo rejsa.'Yesterday morning the child (allegedly) went out to school very early, and yet (allegedly) didn't catch the bus.' The statement in (3a) is appropriate when the speaker has first-hand knowledge of both the going out of the house and the missing of the busimagine a parent who left the house with their child then accompanied them to the bus stop and so witnessed the missing of the bus.The statement in (3b) is appropriate when the speaker has first-hand experience of the leaving of the house, but only indirect knowledge of the missing of the bus -imagine a parent who waved the child good-bye at the house door, but the missing of the bus was reported to them later.And finally, (3c) suggests the parent doesn't have first-hand knowledge of either event, perhaps they had to leave for work before the child left the house to go to school.
Is there, however, uncertainty?Not necessarily.The speaker may be sufficiently committed to the truth of the statements expressed via the evidential to signal, for instance, that they will predicate their own future actions on them.They could, for instance, state that they will change their plans -see (4).
(4) a. Včera sutrinta deteto izleze za učiliste mnogo rano i pak ne xvana rejsa, zatova utre šte ja karam s kolata.'Yesterday morning the child went out to school very early, and yet didn't catch the bus, so tomorrow I will drive her.' b.Včera sutrinta deteto izleze za učiliste mnogo rano i pak ne xvanalo rejsa, zatova utre šte ja karam s kolata.'Yesterday morning the child went out to school very early, and yet (I am told) didn't catch the bus, so tomorrow I will drive her.' c.Včera sutrinta deteto izljazlo za učiliste mnogo rano i pak ne xvanalo rejsa, zatova utre šte ja karam s kolata.'Yesterday morning the child went out to school very early (I am told), and yet (I am told) didn't catch the bus, so tomorrow I will drive her'.
We think this indicates that uncertainty isn't necessarily part of the meaning of the evidential.Is uncertainty a possibility, however?Following up a statement with the indicative tenses with an explicit expression of doubt seems anomalous (5a); (5b) is marginally better, but still sounds odd.
(5) a. #Včera sutrinta deteto izleze za učiliste mnogo rano i pak ne xvana rejsa, no na men ne mi se vjarva.'Yesterday morning the child went out to school very early, and yet didn't catch the bus, but I doubt that.' b. ?Včera sutrinta deteto izleze za učiliste mnogo rano i pak ne xvanalo rejsa, no na men ne mi se vjarva.'Yesterday morning the child went out to school very early, and yet (I am told) didn't catch the bus, but I doubt that.' We conclude that uncertainty is not part of the meaning of the evidential, but is at most a possible pragmatic inference (we can be less certain of the veracity of second-hand information by its very nature).Rather than explicitly signal uncertainty, the evidential suspends the usual, default commitment to truth implied in the Co-operative principle.In Wierzbicka's terms, the formula appropriate to the Bulgarian evidential is the one suggested for Maricopa: 'someone says this/people say this, I don't say: I know this'.It is in this that the evidential is different from epistemic modality, it seems to us.Whereas modality involves modulating commitments to truth, the evidential involves their suspension.
That the evidential itself doesn't necessarily express uncertainty is also clear, we believe, also from the existence of the so-called dubitative, which we discuss below.

Dubitative
The so-called dubitative form formally builds on the evidential.Like the evidential, the dubitative is used by the speaker when they want to suggest that they have no first-hand knowledge of a situation.The dubitative, however, as its label suggests, also conveys the speaker's doubt.Thus, explicit markers of doubt are much more acceptable in the context of the dubitative, cf.(5b) above and ( 6). ( 6) Včera sutrinta deteto izleze za učiliste mnogo rano.Ne bilo xvanalo rejsa, no na men ne mi se vjarva.'Yesterday morning the child went out to school very early.(I am told, apparently) didn't catch the bus, but I doubt that.' Because of this ability to be a marker of doubt, the dubitative is often used in exclamatives ( 7).
(7) a. Njamal bil pari! 'He doesn't have money apparently -as if!' b.Kontuzen bil.Da be, da! 'He's had an injury, they say.I don't believe a word of it!' c.Straxotna bila!Ne, tja e prosto edna pudra.'They say she's great.I don't think so, she's just a big diva.'By using the dubitative, speakers don't simply suspend their commitment to the truth of a statement, they signal that they doubt its truth.In this respect, the dubitative is similar to modality, but it also marks a statement as second-hand knowledge.In terms of NSM, we would suggest a formula that signals both this second-hand knowledge and the lack of the commitment to the truth of the statement, i.e. 'someone else says this, I think this may not be so'.

The inferential
The inferential is related formally to the evidential and/or the perfect, and is used to communicate the speaker's inferences or deductions.As with the evidential, the speaker doesn't have access to first-hand information.According to Târpomanova (Търпоманова, 2015, p. 40), the inferential can be used to report on two types of deductions: (i) a deduction made on the basis of presently available, tangible evidence used to reconstruct past events, or (ii) deduction based on general knowledge, rational argumentation or supposition, knowledge of what normally happens, or inferences of what might happen given certain circumstances.A typical context for the use of the inferential is illustrated in (8).
(8) a. Kradecât se e izkateril po uluka, razbil e prozoreca na vtorija etaži ottam e vljazâl v apartamenta.'(I deduce that) The burglar climbed the drains, broke through the window on the second floor, and entered the apartment from there'.
b. Ekspertiza pokazva, ce šofjorât na avtomobila ot težkata katastrofa v Plovdiv, pri kojato zaginaxa dvama duši, e karal pijan.'The expert report shows that the driver in the Plovdiv car accident which led to the death of two people (it is deduced) was driving under the influence of alcohol'.
As with the evidential, we don't think uncertainty is part of the meaning of the inferential, aside from the normal pragmatic inference that we can be less certain of the veracity of statements that rest on deduction.A speaker could, however, add modality to a statement couched in the inferential, as in ( 9). ( 9) Kradecât moze bi se e izkateril po uluka, razbil e prozoreca na vtorija etaž i ottam e vljazâl v apartamenta.'(I deduce that) The burglar maybe climbed the drains, broke through the window on the second floor, and entered the apartment from there'.
Note that a similar addition of modality to (8b) would entail uncertainty expressed in the medical report, rather than the (implicit) speaker/writer (10).
(10) Ekspertiza pokazva, če šofjorat na avtomobila ot težkata katastrofa v Plovdiv, pri kojato zaginaxa dvama duši, može bi e karal pijan.'The expert report shows that the driver in the Plovdiv car accident which led to the death of two people (it is concluded) was probably driving under the influence of alcohol'.
We suggest that the semantic formula for the inferential is parallel to that of the evidential, although it signals a different source of information: 'I now know this because of things I see/things everyone knows'.

The mirative
And finally, we will venture a semantic formula in the spirit of NSM for the mirative, formally similar to the inferential/dubitative.Although there isn't yet a consensus view on this, we will assume for the time being that the mirative is a particular use of the inferential and/or the dubitative, in which a deduction, an experience, or a perception is accompanied with an expression of stance.Formally similar to the inferential or dubitative, the mirative can be used in situations where the speaker has first-hand knowledge or is making a deduction on the basis of available evidence.Thus (11a) can be said upon seeing the rain, and (11b) can be said whilst walking in nature.However, (11c) would be most appropriate when seeing a prepared dinner and making deductions about its preparation, rather than when witnessing the process of preparation (otherwise there will be no element of surprise), whereas (11d) is equally appropriate when said by someone who has become an education expert, as when said by someone passing judgement/making inferences about someone else being made an education expert.We adapt the formulae for the inferential: 'I didn't know this, now I know this because of things I see/experience, because of this I feel something good/ bad'.As we have tried to convey in this formula, the mirative marks the knowledge of the speaker as new and unexpected.This knowledge-counter-to-expectations causes a positive or negative judgement (the listener must infer from context whether the surprise is negative or positive).

Different contexts, different rules
The overview we offer here isn't exhaustive.We would like to point out one important limitation: most of the examples we offer in the preceding sections come from contexts of past-time narratives.This isn't very surprising: we usually have first-hand experience of situations unfolding in front of our eyes now, but not of all situations that happened in the past, and so it is mostly in these contexts that the distinction between first-hand and second-hand knowledge matters.As we mentioned above, there is a debate in the literature whether the aorist and imperfect involve subjective confirmation on behalf of the speaker.A related but slightly different debate revolves around whether situations reported in the aorist or the imperfect have been 'witnessed' by the speaker.We think one need not include this in the semantics of the aorist or the imperfect.We can assume that they can be used to refer to events that could not have been witnessed by the speaker.However, when recounting events that could have been witnessed but were not, a speaker of Bulgarian has grammatical means at their disposal to indicate this: the evidential.The Co-operative principle (the maxim of quantity) requires that they do so.Therefore, under the assumption that a speaker is observing the principle, failure to use the evidential in such contexts entails that they witnessed the situation they are referring to.
The status of witness/non-witness or, more generally, the status of knowledge as being first-hand or not, is not relevant in all contexts.We can't have first-hand experience of events that happen in the future, for instance.Although this merits additional investigation, we would venture the supposition that in future-time contexts, speakers are less scrupulous about whether they should use evidential forms -in ( 12) the (a), with indicative, and (b), with evidential, examples are equally acceptable.
'According to my daughter, it will rain tomorrow'.
In past narratives, on the other hand, a whole story could be told from the perspective 'of another', i.e. the speaker may choose to use the evidential for events they have first-hand experience of when they are part of one narrative sequence with situations they have no first-hand knowledge of (cf. the walking in the park in 13).
(13) Včera, dokato sme se razxoždali v parka, tja rešila da se obadi na bivšija i da zamine pri nego v Belgrad.'Yesterday, whilst we were (reportedly) walking in the park, she decided (reportedly) to call her ex and go visit him in Belgrade'.

Conclusion
In this paper, we revisit Wierzbicka's account of the evidential in Slavic.Whilst her focus is primarily on Macedonian, we adapt it to Bulgarian data stepping on established accounts of the evidential.We depart from Wierzbicka's analysis in several different ways.This isn't the main thrust of our paper, however.Our disagreement with Wierzbicka's account is only to be expected, given that the view of the Bulgarian verbal system we adopt is different from her reference point.We revisit the NSM account because it places Bulgarian evidentiality in a typological space which is well-defined and described in detail in the rest of Wiezbicka's work.Explicit approaches like NSM in principle provide a solid basis for detailed cross-cultural studies of diverse languages and cultures.
Our other aim is to develop extensions to cover some of the categories related to the evidential.In the sections above, we propose explications in the spirit of NSM of the dubitative, the inferential, and the mirative.Although at this stage, our proposals are tentative, we hope that explicit formulae in an established framework can provide more precise distinctions between these categories.